可於滾筒上製造次微米週期性條紋之系統伺服控制

Abstract

在本論文中，作者設計並實現了一套應用於滾筒上製造次微米級週期性干涉條紋的複雜伺服系統。此系統主要包含了兩大子系統。其一為一個滾筒定位系統，此子系統由數組致動器系統所組成，其中包含了一組高精度線性馬達平台與控制器，一個三軸壓電致動高頻寬精密定位平台，以及一組高精度三軸手動定位平台與一個高解析度DC直流馬達旋轉平台；根據所設計的曝光程序，在滾筒置放於定位系統之前，先以三軸手動平台粗略調整至適當之姿態，透過個人電腦利用C＋＋環境下所開發之程式傳送與接收資料，在滾筒到達曝光位置前由線性馬達平台提供長行程之定位，並由線性平台本身控制器所提供之PID控制器進行定位誤差之修正，而三軸壓電致動平台則用於滾筒傾斜姿態之調整，並由DC直流馬達旋轉平台驅動滾筒，待所有定位程序完成後始進行曝光；文中將說明滾筒定位系統各組成之功能、設計理念及相關討論。 另外一個重要的子系統為量測系統。此系統包含兩組量測儀器：一套具有高頻寬及高精確度的雷射干涉儀量測系統，用以提供滾筒定位系統於X軸向上之位移值、一組雷射位移感測器，透過兩顆感測器進行滾筒傾斜姿態之量測，並回傳至壓電致動平台進行修正補償。 為獲得滾筒定位系統之特性，本研究針對所架構的系統進行鑑別，根據定振幅但頻率變動之弦波速度值輸入與實際量測滾筒之位移值輸出，透過Matlab Toolbox所提供之系統鑑別功能以不同模擬方式進行估測，在多次測試與比較之後得到以State-space形式中之PEM方式模擬可獲得最近似系統模型之結論。 然而在模擬過程中，發現多組在相同狀況下所量測之輸入輸出數據所估測的模型並未有可重複性，可知在此滾筒定位系統中有不確定性因素存在，因此推測在訊號的傳輸間具有jitter (抖動) 現象，故以系統鑑別實驗中最近似之系統模型為基礎，設計jitter現象之相關量測流程以及建立一組可模擬jitter現象對實際系統輸出影響之模型。在量測過程中，根據示波器對線性平台的數位輸出埠進行長時間記錄，經過轉換與運算後顯現出原先應為固定的系統取樣頻率具有不規則之跳動現象，驗證jitter現象存在於本定位系統中，而取樣頻率跳動之範圍亦成為稍後模擬程式中之參數。 模擬程式之概念是採用亂數產生之動態取樣週期與取樣頻率變動範圍當作變數，本論文避開了較複雜而不易觀察結果的方法，採用在考慮不同程度之jitter現象的前提下直接模擬系統的輸出，並引進一個量化指標來粗略地估計jitter程度對實際上系統輸出之直接關係。最後模擬結果呈現此模擬程式之可行性，並驗證其量化指標具有判斷jitter現象嚴重程度對系統輸出影響大小之能力，亦說明具有jitter現象之系統不適用於一般系統鑑別之程序。

In this thesis, we designed and implemented a complex servo system for stitching interference grating patterns on a roller surface. The proposed system included two sub-systems: a “roller positioning system,” which includes a high-accuracy linear motor stage and its controller, and a precision system that is composed of a high-bandwidth and high-precision three-axis PZT stage, a high-precision three-axis manual positioning stage, and a high-resolution DC motor rotary. According to our control strategy, the manual stage would adjust to an adequate position before the roller is put on the positioning system, and then the control program written in Visual C++ environment will transmit and receive the data. The linear motor stage is responsible for the long stroke motion of the roller, and for compensating the positioning error through the PID controller. The posture of roller tilt is compensated by a PZT stage, and then the roller is rotated by a DC rotary motor. The exposure will begin when the positioning procedure was completed. The functions and concepts of the roller positioning system will be discussed in the thesis. Another important sub-system is the measurement system, which includes two measurement instruments. The first is a high-bandwidth and high-accuracy laser interferometer system responsible of measuring the position along X-axis. Second is a set of laser displacement sensors for measuring the roller tilt posture. To obtain the characteristics of the roller positioning system, the system architecture was identified with different simulating methods inside the MATLAB Toolbox. After some tedious estimating and comparing, the simulation result with PEM method in the state-space form can obtain the best system model. The other estimating models with different sets of input/output under the same condition turned out to be not repeatable. It means that there are some uncertainties inside the system, such as jitter phenomenon. We have thus proposed a jitter measurement procedure and simulation model based on the reference system model from the identification. During the measurement, the digital output signal of linear motor stage was recorded by the oscilloscope for a long period and it showed the variations of sampling frequency and proved the jitter phenomenon exists. The concept of the simulation model is based on the random generated dynamic sampling periods and the bound of frequency variations. In this thesis, we avoided the complicated and unobservable method to estimate the system output directly in the consideration of the different order of jitter severity, and introduced a quantification index for rough estimation of the direct relation between system output and jitter phenomenon. Finally, the results show that the feasibility of this simulation model and verify the quantification index can effectively represent the influences of jitter on the system output. They also prove that the general system identification procedure applied on the jitter-affected system did not represent the actual system characteristics.